Live Album Covers AI

Bringing some of Brazil’s most iconic music album covers to life using artificial intelligence.

“Is this sacrilege?”


These were my thoughts once my heart rate returned to normal after the initial shock of seeing the footage of two lively boys (named “Cacau” and “Tonho”) sitting side-by-side on a dirt road. It comprises a few seconds of them staring at one another, the black kid keeping a stern look while the white boy next to him dones a mischievous playful smile. 

Then, they look at each other and back to the camera. 

Original Creations vs. AI Interpretation


Those boys have been part of my life growing up in Brazil, and they live in the collective imagination of millions of Brazilians since the release of the album “Clube da Esquina” in 1972 by Lô Borges and Milton Nascimento – one of the holiest partnerships in the vast pantheon of Brazilian musical geniuses. Now, thanks to artificial intelligence and me, the boys were able to look at one another for the first time. Or this is how they might have moved one day many decades ago, on a country road in Nova Friburgo, had Cafi, the creator of the original picture, made a movie instead of a still image. Back then only a few people witnessed that moment. Now, at the time of me writing this, half a million people have seen them move as well.

With the release of Luma’s Dream Machine, I decided to test the platform’s AI promises of creating unprecedented realistic movies out of thin air, and I proceeded by feeding it with still images of some of Brazil’s most iconic album covers. The goal was to test its technical capabilities but to also continue my ongoing inquiry into how AI systems often misrepresent Brazilian culture. For a few years now,  I have witnessed AI platforms struggling when adding specific Brazilian references to the prompt or even when given exact image references for it to work on top of. When I asked for a “Carnaval parade” it would generate a crowd of southeast asian men or city scenes that, even though looking distinctive Latin American, still had no resemblance to what I see from my window in Rio de Janeiro. Yet when AI churns out a great result -when it gets it “right-, it is like witnessing little revelation. It is as if this amorphous technological system had finally acknowledged my culture’s existence. A feeling that is often followed by the realization that it is also appropriating it – digesting in its corporate algorithmic black box, with unforeseeable future consequences.

A few days ago I published the AI-generated results on Instagram, alongside animated versions of other equally famous covers such as “Secos e Molhados” (1973), Jorge Ben’s “Samba Esquema Novo”(1963), Paulinho da Viola’s “Nervos de Aço”  (1973) and Cartola’s “Verde que te quero Rosa” (1977). For all the wonderment and tears that the boys from “Clube da Esquina” generated, I received horrified responses in the comments section related to the results I generated for the samba mastermind Cartola. It looked weird and grotesque, and I published it anyway, as I decided to show how AI is failing forward. I wanted to show how AI isn’t “intentional”. It has little understanding of context and the unknown, in this case, the face of the artist.

 In the original cover, the Cartola is wearing his eponymous sunglasses and sipping a “cafezinho” that covers half his face and so has he remained still, sipping that coffee for decades. That is, until the publishing of my little Instagram reel, where he suddenly moves the cup out from his face revealing not his famous thin mustache, but a grotesque toothless mouth, babbling and moving incongruently – and devoid of facial hair! In the following seconds, the scene becomes even more surreal, as the cup submerges into the plate and the cigarette between his fingers suddenly gets stuck in his ear. No wonder hundreds of people complained of my ignorance of Cartola and the basic laws of physics. But what should I say? Is it really just the “AI’s fault!”? 

The AI medium is the AI message

Grotesque results are still common when co-authoring content with AI agents. I am aware that Cartola had teeth and that cigarettes don’t float in midair, but I decided to publish the results as a testament to the current state of AI. If the technology keeps evolving at the break-neck pace it has in the last couple of years, and nothing seems to show it will slow down, I believe we will soon look back with nostalgia to this moment in time when AI still made “goofy mistakes”. As strange as Cartola’s mouth might look, it is already a quantum leap compared to the blurred cronembergian mess AI image generators would often give only one year ago. Now, provided with a few words or a single image, systems like Luma’s Dream Machine or Runway ML conjure uncanny realistic moving pictures with real living humans doing complex actions – a few extra fingers or twisting necks notwithstanding. 

Uncanny wonder

The unexpected popular success I glimpsed with this video is a telltale of the widespread wonderment and unease AI-generated content is causing. My answer to most people who say the are “afraid of this thing” is to say “me too…but”. As a creator, the best way to cope with this brave new AI world is to surf it and see where the waves will take me. 

Here’s the full list of Albums depicted in the video.

1. Clube da Esquina (1972) (um dos meus favoritos)

2. Samba Esquema Novo – Jorge Ben (1963)

3. Secos e Molhados (1973)

4. Nervos de Aço – Paulinho da Viola (1973)

5. Verde Que Te Quero Rosa – Cartola (1977)

Share the Post:

available for freelance

Let's Chat

Rolar para cima